AI for CMOs
Your AI needs a brand kit, not just a clever prompt
SimplSolutions editorial team · Brand & creative · 4 min read
Published · Updated
AI-assisted editorial guidance. Jordan is a fictional campaign character, not a verified human author. Worked examples are illustrative unless explicitly identified otherwise.

Your brand kit needs an authority model
An AI-ready brand kit is more than logos and tone adjectives. It tells the writer which facts govern a campaign, who owns them, which details local teams may change and what to do when the information conflicts. Without that authority model, a better prompt can simply produce a more convincing mistake.
The practical deliverable is a source register plus annotated writing examples. This guide explains how to build one for a recurring campaign, test the exceptions and keep it current without spreading more contradictory attachments.
Inventory five kinds of information
Start with service descriptions, offer terms, brand assets, approved voice examples and prohibited claims. Keep the current version findable and identify its owner. A shared folder full of files does not tell a marketer which file is authoritative.
For each source, record its purpose, owner, approved version or date and update path. Do not copy private information into the register. The register points to appropriate sources; it should not become another uncontrolled database.
| Source | Owner decision | Typical failure |
|---|---|---|
| Service description | What the company actually provides | Old or overstated promise |
| Offer terms | Dates, eligibility and conditions | Expired promotion |
| Brand assets | Current approved visual files | Old logo or unsuitable format |
| Voice examples | Why wording fits the company | Vague adjectives with no examples |
| Prohibited claims | What cannot be asserted | Unsupported guarantee |
Split fixed facts from permitted local variation
An approved company promise should remain fixed unless the authorized owner changes it. Addresses, booking links and local availability can vary when verified. Explain those permissions in the template, not just in a policy manual.
For a two-branch campaign, prepare one row for each location with current contact details and eligible service. Include a non-participating branch in the test. The workflow should exclude its offer instead of assuming every office shares the same conditions.
Build voice examples that explain the choice
Show approved and rejected passages with annotations. Identify terminology, amount of detail, evidence and promise strength. "We name the next step clearly" is a usable instruction. "Make it inspiring" leaves too much interpretation.
Keep factual review separate from style. A passage can sound like the company while offering a service that does not exist. Use the brand voice guide to turn examples into acceptance rules.

Illustrative editorial image, not a customer result.
Test a deliberately wrong brief
Use synthetic material containing an expired offer, an unsupported discount and two conflicting descriptions. Ask for a draft and an unresolved-facts note. The correct behavior is to identify the issue and route it to the owner, not quietly choose the most persuasive version.
After clarification, inspect the entire output. Old facts can survive in headings, captions and calls to action even when the body was corrected. Keep source references with the draft where practical so the reviewer can see what it used.
Keep visual assets current without confusing facts
Tools such as Canva Brand Kit can support asset organization. A visual asset library is not by itself the authority for offer terms, service availability or commercial promises. Confirm the actual capabilities and account requirements in current product documentation before relying on them.
Name who removes or retires old files. Retaining an obsolete asset for reference is different from leaving it in the approved-use library. Staff should not have to infer approval from a filename.
Maintain the kit as part of campaign operations
When a source changes, identify affected templates and drafts. Decide which accepted material needs review again. Do not automatically publish a new claim because the source register was updated.
Track corrections by cause: old facts, wrong asset, unclear permission, unsupported claim or tone. If the same issue repeats, change the kit or workflow rather than asking each branch to remember another exception.
Your first useful kit
Create the source register, two annotated voice examples, one local-details table and an exception test for one campaign. Have the owners approve it and run it with two offices before scaling. Get the brand audit worksheet. If you need that approved context available to a reviewed drafting workflow, request a SimplSolutions demo.
